swc-launch only works once per boot

[paulhandy]

using the example wm, i am able to run swc-launch one time each boot. concurrent attempts show it exiting immediately after calling swc-launch ./wm.

[michaelforney]

Hmm, I have not observed this. Does swc-launch attempt to change the VT? That will help identify whether the problem lies in libswc or swc-launch.

[paulhandy]

Yeah, it appears so. it will momentarily go blank for just a moment, but no real meaningful error messages are printed to STDOUT/STDERR.
It does print:
Dropping DRM masterginal state] DEBUG: Adding Device HDA Digital PCBeepTouchPadrd

Edit: maybe the problem is in evdev_devices.c, since that is the only debug message

Edit2: actually, it seems that the stdout is getting all mucked about. Dropping DRM Master got printed on its own line one time and everything else got printed clearly on their own lines. I have no idea what would be causing interrupts like that.

[michaelforney]

It sounds like the example wm might be crashing. You might be able to figure out why with swc-launch -- valgrind ./wm or swc-launch -- gdb -batch -ex run ./wm

[paulhandy]

with valgrind, it complains that ./wm does not exist.

with gdb -batch -ex run, it booted it up, but with the following warnings:
warning: Could not load shared library symbols for linux-vdso.so.1
Do you need "set solib-search-path" or "set sysroot"?

Why would it work with gdb and not normally? or is gdb normal?

[crondog]

So I am having the same issue with the example wm and velox (nice work btw). I tried running with valgrind and gdb but it does not crash. However if i just redirect stderr i can get some nice output. I hope this helps

swc-launch -- ~/velox/velox 2> /tmp/swc.log 

[swc:libswc/drm.c:186] DEBUG: /dev/dri/card0 is the primary GPU
# find_driver: Trying DRM driver `intel'
# find_driver: Trying DRM driver `nouveau'
# find_driver: Trying DRM driver `dumb'
Using config at '/etc/velox.conf'
Server exited with status 0
Restoring VT to original state
Dropping DRM master

[jezze]

I don't know if this is related but I'm having a trouble with a race condition happening when the launcher is trying to setup tty1 which gives the same result. It is between the ioctl call for VT_ACTIVE and VT_WAITACTIVE that this sometimes happen. A quick fix is to insert a sleep(1) between the two calls.

[crondog]

@jezze So using sleep(1) fixes that issue for me. It also allows me to run swc-launch in another tty which I wasn't able to do before either :)

Now off to investigate further

[michaelforney]

Can anyone check if this is still an issue?

[logankaser]

@michaelforney I'm still getting this on the august 30 commit; I'll update later today and test again.